Diffuse soil CO 2 degassing from Linosa island

Herein, we present and discuss the result of 148 measurements of soil CO2 flux performed for the first time in Linosa island (Sicily Channel, Italy), a Plio-Pleistocene volcanic complex no longer active but still of interest owing to its location within a seismically active portion of the Sicily Channel rift system. The main purpose of this survey was to assess the occurrence of CO2 soil degassing, and compare flux estimations from this island with data of soil degassing from worldwide active volcanic as well as non-volcanic areas. To this aim soil CO2 fluxes were measured over a surface of about 4.2 km2 covering ~80% of the island. The soil CO2 degassing was observed to be mainly concentrated in the eastern part of the island likely due to volcano-tectonic lineaments, the presence of which is in good agreement with the known predominant regional faults system. Then, the collected data were interpreted using sequential Gaussian simulation that allowed estimating the total CO2 emissions of the island. Results show low levels of CO2 emissions from the soil of the island (~55 ton d-1) compared with CO2 emissions of currently active volcanic areas, such as Miyakejima (Japan) and Vulcano (Italy). Results from this study suggest that soil degassing in Linosa is mainly fed by superficial organic activity with a moderate contribution of a deep CO2 likely driven by NWSE trending active tectonic structures in the eastern part of the island.


Introduction
The release of gases such as CO 2 from the soil is a common process that occurs in both tectonic and volcano-tectonic areas.In active volcanic areas, large amounts of CO 2 are released from ascending magmas or generated by crystallization of deep magma bodies [Harris andRose 1996, Camarda et al. 2012].Part of this CO 2 is released into the atmosphere through summit craters, while another part is emitted from the soil along the flanks of volcanoes.In non-volcanic environments, moderate to strong emission of CO 2 are usually encountered in areas with the occurrence of active tectonic structures, such as faults and fractures, which constitutes highly permeable preferential pathways for natural emissions of deep gases.Because tectonic lineaments are often buried by soil and/or sediments that act as a gas diffuser, measurements of gas emissions from soil can provide useful information regarding the presence of hidden and/or active faults [Giammanco et al. 1997, Ciotoli et al. 1998, Lewicki and Brantley 2000, Guerra and Lombardi 2001, De Gregorio et al. 2002, Aiuppa et al. 2004, Giammanco et al. 2006], as well as changes in the in feeding system of active volcanic areas [Giammanco et al. 1995, Federico et al. 2011, Camarda et al. 2012].
Here we discuss the results of a soil CO 2 flux survey performed in Linosa island.The island is located in the south-western part of the Mediterranean Sea (see Figure 1), and with Pantelleria island constitutes part of the Sicily Channel rift system (SCrs).Volcanism occurred at Linosa island in the late Plio-Pleistocene [Lodolo et al. 2012, and references therein], while there is no evidence of current volcanic activity, such as fumaroles or thermal waters.Only few submarine exhalative manifestations were identified near the top of a submerged relict volcanic structure located in the SW offshore of the island [Lanti et al. 1988].
The main aim of this study was to investigate the soil CO 2 flux in a volcanic island with no signs of a recent volcanic activity but located in a currently active seismic area.The results of this study can contribute to extend our knowledge of the global CO 2 emissions within volcanic and non-volcanic areas [Burton et al. 2013].

Regional and local tectonic setting
The Sicily Channel rift system (Figure 1) formed as consequence of the continental rifting process that affected the northern margin of the African plate since the Late Miocene in response to the opening of the Tyrrhenian Sea [Boccaletti et al. 1987].It is characterized by three main tectonic depressions known as Pan-

Subject classification:
Fluid geochemistry, CO 2 , Soil degassing, Volcano monitoring, Instruments and techniques.telleria, Linosa and Malta grabens with depths of 1350, 1580 and 1720 m below sea level (b.s.l.), respectively [Grasso et al. 1991, Civile et al. 2010].The uprising of the asthenosphere to 60 km in depth [Della Vedova et al. 1989], and the thinning of the African foreland continental crust [Scarascia et al. 2000] resulted in the formation of widespread volcanism that characterized the Sicily Channel.The formation of the short-lived island of Ferdinandea in 1831 from the Graham Bank area [Gemmellaro 1831] and the submarine eruption of the volcano Foerstner in 1891 [Washington 1909] sited 5 km NW of Pantelleria island, were described as the last two volcanic events responsible of a wide range of products from basalt to trachytic/pantelleritic composition with different geochemical affinity, such as tholeiitic, alkaline, and peralkaline [Rotolo et al. 2006].Thermal and bubbling springs, mofettes and fumaroles are all present in Pantelleria island.In addition, the actual seismic activity [Schorlemmer et al. 2010] is recorded principally near the axial zone of the Sicily Channel (Linosa and Pantelleria Grabens).
Located about 160 km south of the southern Sicilian coast in the Pelagian Archipelago (Figure 1), Linosa island (about 5.2 km 2 ; 196 m a.s.l.) represents the subaerial portion of a larger volcanic complex rising from a depth of 800 m b.s.l.near the south-west sector of the homonymous graben [Calanchi et al. 1989].The influence of the regional tectonic trends (NW-SE and NNW-SSE) on the volcanic evolution of the island can be inferred both by the distribution of the inland and off-shore volcanic centers (i.e. the ancient submerged volcanic center of Secca di Tramontana and a shoal in the SE margin), and by the maximum axis of elongation of some principal edifices, such as Montagna Rossa and Pozzo Salito (Figure 2).These edifices represent the most ancient sub-aerial volcanic complexes in the SCrs dated 1.06 ± 0.10 and 0.53 ± 0.07 Ma [Lanzafame et al. 1994].

Materials and methods
We performed 148 measurements of soil CO 2 flux during the summer in 2006.The measurement points were distributed over a surface of about 4.2 km 2 (80% of the entire area) passing through all the main eruptive centers of the island (Monte Vulcano, Montagna Rossa, Monte Nero, Timpone, Monte Bancarella, Fossa Cappellano and Pozzo Salito; see Figure 2), with the exception of some lava fields located along the NW and SW side of the island (Caletta and Arena Bianca).In these last areas the soil was found to be totally absent or not thick enough to perform CO 2 flux measurement with the method described hereinafter.The soil gas investigation was carried out with a mean distance of 150 m between adjacent measurement points, chosen as a good compromise between the detail of the survey and the extent of the investigated area.It is known that the diffuse degassing from soil is influenced by atmospheric conditions [Diliberto et al. 2002, Granieri et al. 2003, Rinaldi et al. 2012].To minimize the effect of the variability of atmospheric conditions the measurements were conducted within few hours during the day, implying a negligible temperature variation.
Measurements of CO 2 diffuse degassing were performed using the dynamic concentration method [Gurrieri andValenza 1988, Camarda et al. 2006b].This method consists of measuring the CO 2 concentration in a mixture of air and soil gas that has been generated in a specifically designed probe inserted into the soil to a depth of ~50 cm.The gas mixture is then obtained by producing a very small negative pressure in the probe using a pump at a constant flux of 0.8 L min -1 .The mixture is subsequently analyzed using an infrared (IR) Gascard II spectrophotometer provided by Edinburgh Instruments Ltd.After a given time of pump activation (generally less than 1 minute), a constant CO 2 concentration is measured in the gas mixture, that is the dynamic concentration C d [Camarda et al. 2006b].The relationship used to convert dynamic concentration values to CO 2 flux was calibrated after laboratory measurements during which several measurements of C d were collected in a soil layer fed by known and constant CO 2 fluxes [Camarda et al. 2006b].Several tests were performed using soils with different values of gas permeability and porosity in order to evaluate the influence of soil permeability on the calculated CO 2 fluxes.These results revealed that the propagated error in the flux calculation is generally less than 5% for soils with gas permeability varying between 0.36 and 123 darcy [Camarda et al. 2006b].
Due to the absence of values of gas permeability for the soils of Linosa, we considered a value of soil gas permeability of 37 darcy used to calculate the CO 2 flux from all the 148 measurement points with dynamic method.This value corresponds to the average value of gas permeability estimated for the soils in Vulcano island, Sicily [Camarda et al. 2006a].The soil gas permeability in this island ranges between 5 darcy (fine sand) and 80 darcy (coarse sand).Other values of gas permeability of volcanic soils were reported by Moldrup et al. [2003], where the air permeability of some volcanic ash soils were determined from three locations in Japan.In this case the reported values ranged from 1.8 to 168 darcy with a mean value of 31 darcy.Additional data available in literature generally refer to water permeability and cannot be used to accurately assess the value of gas permeability of volcanic soils due to the known Klinkenberg effect [Klinkenberg 1941].Therefore, our value of 37 darcy can be considered representative for gas permeability of soils in Linosa island.
Further uncertainties in measurements of CO 2 from soil using the dynamic concentration method might be introduced by pressure variations during the field measurements.However, owing to the almost flat topography of the island (max altitude of ~190 meters), possible errors can be considered negligible (about ±3% of range).
During the survey three gaseous samples were collected in those sites with the highest CO 2 fluxes to determine their carbon isotopic composition.Gases were DIFFUSE SOIL CO 2 DEGASSING FROM LINOSA ISLAND sampled from the soil at a depth of 50 cm using a Teflon tube of 5 mm in diameter connected to a syringe, and then stored in glass flasks equipped with vacuum stopcocks.The carbon isotopic composition of CO 2 was determined using a Finnigan Mat Delta Plus Mass Spectrometer.Measured isotopic values are expressed as d 13 C(CO 2 ) versus the Pee Dee Belemnite (PDB) standard and characterized by uncertainties of ±0.2‰.

Sequential Gaussian simulation and estimation of total CO 2 flux
Sequential Gaussian Simulation (SGSIM) was performed with the aim to map the spatial variation of soil degassing and calculate the total CO 2 output of the island.
The (SGSIM) was performed using the Stanford Geostatistical Modeling Software (SGeMS), an opensource package developed by the Stanford Center for Reservoir Forecasting for solving problems involving spatially related variable [Remy 2005].To perform SGSIM the distribution of data must be Gaussian.Therefore, as first step, all fluxes were normalized in order to obtain a distribution with mean and variance being equal to 0 and 1, respectively.Then, the semivariogram of normalized data was determined in order to examine the spatial variability of the measured soil CO 2 fluxes.
Further technical details of the SGSIM consisted of, (1) definition of the grid of the simulation; (2) definition of a random path visiting each node of the grid area; (3) random choice of the node in the sampled grid; (4) estimation at each node of the mean value and the variance of the local conditional cumulative distribution function (CCDF) by using simple kriging algorithm considering the semivariogram of the normalized data; (5) drawing of a value from that Gaussian CCDF and addition of the simulated value to the data set; (6) proceeding to the next node in the path and repeating the (4) and ( 5) steps until all nodes have been visited.
The total CO 2 output emitted from the soils of the island of Linosa and its relative uncertainty were obtained as the mean and the standard deviation of the total soil CO 2 flux estimated for each of the 100 simulations.These flux values were calculated for each simulation integrating over the grid area the average value of the soil CO 2 flux given by the SGSIM.

Statistical analysis of the soil CO 2 fluxes
The data relative to the measured dynamic concentrations (C d ) and the calculated fluxes are shown in Table S1 along with the coordinates of each measurement point (see Supplementary files).The values of soil CO 2 flux collected during the survey showed a small variation between 12 and 78 g m -2 d -1 , with a mean value of 11 g m -2 d -1 .This range of fluxes is lower than that observed in active volcanic areas, such as the island of Vulcano and Mount Etna (Italy), where the measured soil CO 2 fluxes generally display values up to three orders of magnitude [Chiodini et al. 1998, Camarda et al. 2012].The high variability of soil CO 2 fluxes observed in active volcanic areas can be manly explained as due to mixing between CO 2 with different origin such as (a) microbial decomposition of soil organic matter and root respiration, and (b) emissions from a deep degassing source [Camarda et al. 2007, Chiodini et al. 2008].As a consequence, the resulting statistical distribution is expected to be, at least, bimodal.
In this study, although the observed low variability of soil CO 2 fluxes, we constructed a normal probability plot (PP) of the soil CO 2 fluxes at the aim to discriminate populations of data ascribable to different sources [Sinclair 1974, Chiodini et al. 2008].This type of plot shows the cumulative frequencies (X axis) of the measured soil CO 2 fluxes (Y axis; Figure 3) on a probability scale.Data that are normally distributed will appear in this plot as a straight line.The presence of multiple straight lines with different slopes indicates, therefore, the presence of various populations of data with normal distribution but with diverse ori-  The main statistics of these populations are listed in Table 1.
To test the validity of the three-population model we also compared the theoretical distribution resulting from the combination of the portioned populations (dotted line in Figure 3) with the actual data distribution (empty circles in Figure 3).A good agreement between the real and ideal distributions was found, as proof of the high goodness of the three-population model.In particular, the population I is characterized by low fluxes with a mean value of 9 g m -2 d -1 and a standard deviation of 1 g m -2 d -1 .These low values are consistent with typical values reported for superficial soil respiration (3.6-13.7 g m -2 d -1 , Monteith et al. [1964]; 6-11 g m -2 d -1 , Brown and Rosenberg [1971]).
The population III is characterized by a mean value of 43 g m -2 d -1 and standard deviation of 21 g m -2 d -1 , and can be considered as the anomalous population, likely suggesting the possible contribution of a deeper source of CO 2 .The intermediate population II shows a mean value of 13 g m -2 d -1 with a standard deviation of 2 g m -2 d -1 .This flux group can be explained assuming mixing of shallower biogenic CO 2 and CO 2 of deep origin [Chiodini et al. 2008].Alternatively, the population II can also be ascribed to the biogenic superficial activity, being its flux consistent with those values typically reported for soil respiration.In this case, we would have two different background populations (I and II) that, for instance, may be related to the presence of different types of vegetation in the island.

Mapping soil CO 2 flux and total CO 2 output
The experimental semivariogram of the normalized data (Figure 4) has been computed and fitted by the following standardized exponential model,  The values assumed by the nugget (C 0 ), the sill (C) and the range (a) are reported in Figure 4 and are here used in the SGeMS software to adjust the model of the SGSIM.
Figure 5 shows three examples of maps representative of low, intermediate and high flux, respectively (A, B and C).The spatial distribution of soil CO 2 flux is significantly different for the three cases.Areas with high flux resulting from one simulation may have lower flux resulting from another simulation.However, the different simulations are equiprobable and "honor" the experimental data in the sense described by Journel and Alabert [1989].Indeed, each simulation has the same univariate statistics and variogram as the experimental data, and furthermore retains the measured flux value unchanged at the original measurement sites.Therefore, we cannot use a single simulation to trace degassing tectonic structures in Linosa island.
In order to show a more suitable map and define the extent of the anomalous degassing areas of Linosa island, as demonstrated by Cardellini et al. [2003], we elaborated a probability map that displays the probability that the soil CO 2 flux exceeds a threshold value at each node of the grid.Figure 6 shows a map with probability exceeding 18 g m -2 d -1 .This value was selected by inspecting the probability plot of Figure 3, and it was assumed as cutoff value separating the intermediate CO 2 flux from anomalous values.As discussed by Cardellini et al. [2003] the map in Figure 6 represents the probability that each location belongs to Diffuse Degassing Structures (DDS).These are well-de-fined regions of a surveyed area characterized by high levels of soil CO 2 emission and they are generally related to structural discontinuities acting as preferential pathways for gas migration towards the surface [Chiodini et al. 2001, Cardellini et al. 2003].The map shows that in Linosa island, DDS lie manly in the eastern sector of the island and have generally a preferential NW-SE direction of elongation, corresponding to the main direction of the faults system in the Sicily Channel.The distribution of the low-energy seismic events along the axial portion of the rift system or in correspondence of the graben borders [Agius and Galea 2011], would confirm the activity of these faults.Hence, the persistent seismic activity in the SCrs would contribute to maintain a high permeability of soils driving deep fluids towards the surface.
The carbon isotopic composition of the CO 2 of three gas samples collected at 50 cm deep in the soil from two different areas showed d 13 C(CO 2 ) values versus PDB of −17.3, −14 and −8.4 ‰, respectively.Due to the low values of the soil CO 2 flux (< 80 g m -2 d -1 ), the kinetic isotopic fractionation can be considered negligible in any case [Camarda et al. 2007].The measured d 13 C values fall within the typical range of CO 2 with organic origin (−30 to −10 ‰ vs. PDB; Hoefs [1980], O 'Learly [1988]), except for the sample with d 13 C(CO 2 ) of −8.4 ‰.
There are two possible explanations to justify the observed isotopic values.As first, the observed variability can be ascribed to different types of plants present in the island.For example, C 3 plants are characterized Figure 6.Map that probability exceeds the threshold value of 18 g m -2 d -1 , discriminating background from anomalous CO 2 flux.The main volcano-tectonic lineaments (red lines) by Lanzafame et al. [1994] are also reported.by d 13 C(CO 2 ) of −27 ‰ whereas C 4 plants have value of −13 ‰ [Cheng 1996].An alternative explanation that can explain the observed data is that the observed shift toward less negative values of d 13 C might be due to contribution of CO 2 from a deep (mantle) source that reaches the surface through the mapped DDS [Frezzotti et al. 2009].However, our data do not allow us to rule out with accuracy one of these two mechanisms, and both might also coexist.
Table 2 reports the total CO 2 flux output calculated as explained above (Section 3.2).In the same table we reported for comparison the total CO 2 output estimated from other authors from some active volcanoes, tectonic and non volcanic areas in the world.As ex-DIFFUSE SOIL CO 2 DEGASSING FROM LINOSA ISLAND pected, the total CO 2 flux estimated from this study for Linosa island is low as compared with active volcanic areas, also considering the much wider areas under investigation.For instance, the flux calculated by Inguaggiato et al. [2012] in the island of Vulcano over a comparable surface, resulted to be one order of magnitude higher than the value found in Linosa.Indeed, as shown here, soil CO 2 fluxes in Linosa are only fed by superficial organic activity and degassing of mantle CO 2 through the recognized DDS.No CO 2 contribution from a hydrothermal systems, such as that identified in Vulcano island [Nuccio et al. 1999], can be claimed in case of Linosa island.Furthermore, by inspecting the CO 2 emission standardized for area (fourth column in Table 2), Linosa resulted to be one of the areas with the lowest CO 2 emission in the world.Nevertheless, his standardized value is comparable with the CO 2 emitted from the Cuicocha and Pululahua caldera in Equador [Padròn et al. 2008].

Conclusions
A survey of soil CO 2 diffuse degassing in Linosa island was conducted during summer 2006, over ~80% of the total surface.The island represents the oldest sub-aerial volcanic centre in the Sicily Channel (1.06 to 0.53 Ma), and to date direct evidences of related volcanic manifestations (fumaroles, thermal springs) are lacking.The results from soil CO 2 flux measurements would confirm this current state of the island.The Sequential Gaussian simulation allowed estimating total CO 2 output from the island of Linosa, which appears very low in comparison to active volcanoes and other inactive volcanic areas.In conclusion, our data showed that the actual soil CO 2 degassing in Linosa is primarily fed by CO 2 with superficial organic (biogenic) origin with a moderate contribution limited to some areas where the origin of CO 2 fluxes can be likely referred to deeper sources.By using a map of the probability that CO 2 flux would exceed a cutoff value we identified the location and the extent of the diffuse degassing structures in the eastern sector of the island.Carbon isotopic composition of CO 2 from samples collected in relatively high-flux zones would suggest a moderate contribution of CO 2 with mantle origin.However, we do not exclude the possibility of a more rigorous geochemical characterization of these gas emissions in future aimed to confirm the possibility of a degassing mantle source beneath the island of Linosa.

Figure 2 .
Figure 2. Simplified representation of the island of Linosa with the main tectonic lineaments (black lines) and volcanic centers (grey areas and dotted lines) (modified after Lanzafame et al. [1994]).Black points indicate the location of measurements for the soil CO 2 flux.

Figure 3 .
Figure 3. Probability plot of soil CO 2 fluxes measured on the island.The red dotted curve shows the theoretical distribution produced by combining the populations in the proportions indicated by the threshold values according to the equation reported at the top of the graph.Solid lines indicate the populations reconstructed according to the procedure illustrated by Sinclair [1974].

Figure 4 .
Figure 4. Variogram of normal scores of the soil CO 2 flux.

Figure 5 .
Figure 5. Maps of soil CO 2 flux.From the top to the bottom, the map relative to the realization giving the lowest value of the soil CO 2 flux output, the map relative to the realization giving the mean value of the soil CO 2 flux output, and the map relative to the realization giving the highest value of the soil CO 2 flux output.

Table 1 .
Summary statistics for the portioned CO 2 flux populations.

Table 2 .
Total CO 2 flux output estimated for the island of Linosa compared with those of some active volcanoes and those of tectonic, hydrothermal or inactive volcanic areas; data are sorted based on CO 2 output in ton km -2 d -1 .